Sensors and Neural Networks to Supervise Integrated Renewable Energy Systems (IRES)
Abstract
The purpose of the study was to consider a sensor-neural- network control system as a means to supervise integrated renewable energy systems (IRES) and to automate their operation. This work suggests an approach to assess the information provided by a set of sensors strategically located in the IRES and help generate control signals in discrete events of time to actuate controllers to best meet the energy and other needs of a small rural area, using the locally available renewable resources such as biomass, insolation, wind, and hydro. The key is to match the resources and the needs a-priori and make decisions based on a prioritized set of needs. Neural networks process the sensor outputs to perform this function of prioritization and generate appropriate control signals. This work uses layered recurrent neural network architecture and is trained using backpropagation. All simulations are performed using MATLAB software. For simulation purposes, a small rural area is considered. Neural networks are trained for four practical scenarios of resource availabilities. The main advantage of using neural networks is that it eliminates the need for complex computations involved in resource need-matching. Results show that in each scenario, appropriate control signals are generated using neural networks. This model is ideal for employing IRES at remote rural locations with minimal investment to best utilize all the available resources efficiently and energize such areas.
Collections
- OSU Theses [15752]